netdev - Re: [PATCH v3 6/6] net: sh

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-id: <4FB1DFF0.4040709@renesas.com>
Date:	Tue, 15 May 2012 13:47:44 +0900
From:	"Shimoda, Yoshihiro" <yoshihiro.shimoda.uh@...esas.com>
To:	David Miller <davem@...emloft.net>
Cc:	netdev@...r.kernel.org, linux-sh@...r.kernel.org
Subject: Re: [PATCH v3 6/6] net: sh_eth: use NAPI

2012/05/15 7:50, David Miller wrote:
> From: "Shimoda, Yoshihiro" <yoshihiro.shimoda.uh@...esas.com>
> Date: Mon, 14 May 2012 15:47:24 +0900
> 
>> This patch modifies the driver to use NAPI.
>>
>> Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@...esas.com>
> 
> I think your TX path is still extremely racey.
> 
> No locks are held here, so you tell me what happens if we execute:
> 
>> +	/* check txdesc */
>> +	txfree_num = sh_eth_txfree(ndev);
>> +	if (txfree_num)
> 

Thank you for your review.
In the sh_eth_txfree(), it check the tx descriptors. If the tx descriptor
is completed, it calls dev_kfree_skb_irq(), and it returns value 1 or more.
So, the sh_eth_poll() calls netif_wake_queue() if the dev_kfree_skb_irq() is
called.

> and at this exact moment the queue was in fact already awake and
> another thread of control transmits packets, and this action fills up
> the TX queue and stops the queue.
> 
>>  		netif_wake_queue(ndev);
> 
> This will erroneously wake the queue and trigger the debugging
> message in your TX function.
> 
> You need strict synchronization between your TX queueing and TX
> liberation flows.  So that queue stop and wake are only performed
> at the correct moment.

I will add netif_queue_stopped() in the sh_eth_poll().

> In fact, looking at how the mdp->lock is used in your TX routine, it
> seems to protect absolutely against nothing.

I wlll remove the mdp->lock in the sh_eth_start_xmit().

> Please read the TX flow of drivers/net/ethernet/broadcom/tg3.c to see
> how to do this correctly, and lock free, in a NAPI driver.
> 

Thank you for your suggestion.
I will add netif_tx_lock/unlock before and after netif_wake_queue().

Best regards,
Yoshihiro Shimoda
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html